next-20090107: WARNING: at kernel/sched.c:4435
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 63 lines · Ingo Molnar
Quoted 48 lines · Alexey Zaytsev
Quoted 40 lines · Ingo Molnar
Quoted 37 lines · Alexey Zaytsev
Quoted 20 lines · Nick Piggin
Quoted 8 lines · Ingo Molnar
Quoted 2 lines · Alexey Zaytsev
One more instance of http://marc.info/?l=linux-kernel&m=123134586202636&w=2
Added Ingo Molnar to CC.added Nick on Cc:. Nick, it's about:
Quoted 7 lines
Quoted 5 lines
commit
7317d7b87edb41a9135e30be1ec3f7ef817c53ddAuthor: Nick Piggin <[email protected]> Date: Tue Sep 30 20:50:27 2008 +1000sched: improve preempt debugging
causing a seemingly spurious warning.
I don't know how it is spurious... Presumably the sequence _would_ have caused preempt count to go negative if the bkl were not held...
Quoted 8 lines
__do_softirqdoes a__local_bh_disableon entry, and it seems like the_local_bh_enableon exit is what causes this warning. So something is unbalanced somehow. Or is it some weird thing we do in early boot that I am missing?Can you put in some printks around these functions in early boot to get an idea of what
preempt_countis doing?Hi again.
Finally got to debug this. The preempt count on the first
__do_softirqentry ever is 0, as it is set inirq_ctx_init(). The interrupted swapper thread happens
to be in thekernel_locked()state at the moment, so the warning.Quoted 6 lines
Quoted 2 lines
I don't understand why the softirq preempt count is initialized to 0. Should not it be
SOFTIRQ_OFFSETinstead?hm, indeed. So this triggers on irqstacks, if an irq happens to hit the first time a softirq executes (ever)? After that point the
preempt_countin the irq-stack ought to stay elevated.Quoted 11 lines
No, this happens on the first softirq, which is run after an irq. An irq interrupts the swapper thread while it is holding the blk. It is executed on the hard irq stack, and the corresponding thread_info.preempt_count is set correctly by
irq_ctx_init(), so nothing happens. After the hard IRQ is over, a softirq is run on the soft irq stack, butirq_ctx_init()set it'spreempt_countto zero. So after the first softirq os over,sub_preempt_count()discovers that the preempt count is goind back to zero, while the BKL is held (by the interrupted thread), and refuses to decrease the count. So the spftirqpreempt_countstaysSOFTIRQ_OFFSETwhich is now correct, so no further warnings are triggered.
Quoted 1 line
yeah. So we need to fix the initial softirq-stack
preempt_countvalue.
Like this? ;)
From: Alexey Zaytsev <[email protected]>
Subject: [PATCH] Set the initial softirq preempt count to SOFTIRQ_OFFSET
Does not changes the preemption semantics, as the softirq's obviously can't be preempted, but fixes a spurious warning in sub_preempt_count, which happens when the preempt count is returned to zero, and the interrupted thread is holding the BKL.
---
arch/x86/kernel/irq_32.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
arch/x86/kernel/irq_32.c +1 -1
@@ -141,7 +141,7 @@ void __cpuinit irq_ctx_init(int cpu) irqctx->tinfo.task =NULL ; irqctx->tinfo.exec_domain =NULL ; irqctx->tinfo.cpu = cpu; - irqctx->tinfo.preempt_count =0 ; + irqctx->tinfo.preempt_count =SOFTIRQ_OFFSET ; irqctx->tinfo.addr_limit =MAKE_MM_SEG (0 ); softirq_ctx[cpu] = irqctx;
Ingo Molnar
AppliedPATCH 1/1 sched: improve preempt debugging
Quoted 11 lines · Alexey Zaytsev
From: Alexey Zaytsev <[email protected]>
Subject: [PATCH] Set the initial softirq preempt count toSOFTIRQ_OFFSETDoes not changes the preemption semantics, as the softirq's obviously can't be preempted, but fixes a spurious warning in
sub_preempt_count, which happens when the preempt count is returned to zero, and the interrupted thread is holding the BKL.Signed-off-by Alexey Zaytsev <[email protected]>
applied to tip/sched/urgent, thanks Alexey.
I also re-added Nick's original patch on top of that (see it below) - this way we'll see whether it's all sorted out for good.
Thanks,
Ingo
------------------>
Quoted 1 line
From
ed094285f3495ad5393fc424011c926f1b6138c1Mon Sep 17 00:00:00 2001
From: Nick Piggin <[email protected]>
Date: Tue, 30 Sep 2008 20:50:27 +1000
Subject: [PATCH] sched: improve preempt debugging
This patch helped me out with a problem I recently had....
Basically, when the kernel lock is held, then preempt_count underflow does not get detected until it is released which may be a long time (and arbitrarily, eg at different points it may be rescheduled). If the bkl is released at schedule, the resulting output is actually fairly cryptic...
With any other lock that elevates preempt_count, it is illegal to schedule under it (which would get found pretty quickly). bkl allows scheduling with preempt_count elevated, which makes underflows hard to debug.
[ This is a re-try of this commit:
7317d7b: sched: improve preempt debugging
which was reverted in:
01e3eb8: Revert "sched: improve preempt debugging"
because this commit:
d04d99d: x86: set the initial softirq preempt count to SOFTIRQ_OFFSET
Fixes the real cause of the spurious warning. ]
--- kernel/sched.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
kernel/sched.c +1 -1
@@ -4440,7 +4440,7 @@ void __kprobes sub_preempt_count(int val)/* * Underflow? */ -if (DEBUG_LOCKS_WARN_ON (val >preempt_count ())) +if (DEBUG_LOCKS_WARN_ON (val >preempt_count () - (!!kernel_locked ())))return ;/* * Is the spinlock portion underflowing?